PDA

View Full Version : Anyone notice Gedmatch Genesis one-to-one matches have really low SNPs?



therrien.joel
10-11-2017, 01:42 AM
I've seen quite a few matches on there where the reported cM are fairly large, up to 10cM but only have 200-300 SNPs. This is even for matches between the same testing company (23andMe in my case). Anyone else notice this? I'm just wondering if that means the matches should be considered less reliable?

Peter Towey
10-14-2017, 12:58 PM
I've noted that too. Also, I have three kits on Genesis: Living DNA, ancestry & FTDNA. I am finding a lot of 23andMe kits arriving as close matches in the last week. With some, my matches on my Living DNA kit are quite high (80cMs or 65cMs) but to one of the others, they are matching at 7cMs and the other, not at all.
I know that Genesis is a BETA site, but this seems very odd! Does anyone have any better info?

Ann Turner
10-14-2017, 11:10 PM
The SNP overlap between some chips is very low, which makes it hard to achieve the "standard" SNP threshold.

https://isogg.org/wiki/Autosomal_SNP_comparison_chart

GEDmatch Genesis tries to adjust the threshold, but I think it's going to be very tricky. My current rule of thumb is to consider only matches of 20 cM or more [edit to add] for the longest segment.

firemonkey
10-15-2017, 12:59 AM
The SNP overlap between some chips is very low, which makes it hard to achieve the "standard" SNP threshold.

https://isogg.org/wiki/Autosomal_SNP_comparison_chart

GEDmatch Genesis tries to adjust the threshold, but I think it's going to be very tricky. My current rule of thumb is to consider only matches of 20 cM or more.

Is that 20cM total or longest block?

Ann Turner
10-15-2017, 04:14 PM
Longest block. I'll edit my post to make that clear.

therrien.joel
10-15-2017, 10:59 PM
I did get an answer from Curtis Rogers one of the developers from Gedmatch.


The Genesis algorithm has been developed to show matches between kits that don't have the same mix of SNPs. That results in less "overlap" than you would see in a comparison between 2 kits from the same testing company, and chip. Because of the new algorithm, the number of SNPs in the matching segment has less importance. We still feel like the 7cM threshold is important, but we see good results with as little as 200 SNPs in the segment. As we learn more about the capabilities of Genesis, we may need to adjust that line of thinking, but in general lower numbers of SNPs in a segment are OK.

Other than cM size of a matching segment, you should also look at the "overlap" figure. I don't have any hard numbers for guidance, but in general, I would be wary of overlaps below about 150,000 to 200,000 . Overlaps above 400,000 are probably an indication of high reliability.

As for the 20cM segment length. That is a very safe choice. Sadly, some of us trying to chase down more distant connections don't have that luxury. I've been fortunate though to find a few cases of phased kits to compare and will try to have that done on my own if I can convince my folks to test. Otherwise I have also succeeded in finding one potential distant relative where I could do triangulation. I guess it's a case of we do what we can with what we have!

That having been said, a closer relative I found through DNA contributed an interesting tidbit that made me change my flags. Turns out my Finnish great-grandmother actually emigrated from the Russian Empire. Still she is clearly at least partly Finnish ethnically, but I can't say she is completely. Go figure, Anthrogenica actually has a Russian imperial flag :)